Concise Review: Patient-Derived Olfactory Stem Cells: New Models for Brain Diseases§


  • Alan Mackay-Sim

    Corresponding author
    1. National Centre for Adult Stem Cell Research, Eskitis Institute for Cell and Molecular Therapies, Griffith University, Brisbane, Queensland, Australia
    • National Centre for Adult Stem Cell Research, Eskitis Institute for Cell and Molecular Therapies, Griffith University, Brisbane, Queensland 4111, Australia
    Search for more papers by this author
    • Telephone: +61-(0)-7-3735-7563; Fax: +61-(0)-7-3735-4255

  • Author contribution: Alan Mackay-Sim conceived, wrote and edited the manuscript.

  • Disclosure of potential conflicts of interest is found at the end of this article.

  • §

    First published online in STEM CELLSEXPRESS September 7, 2012.


Traditional models of brain diseases have had limited success in driving candidate drugs into successful clinical translation. This has resulted in large international pharmaceutical companies moving out of neuroscience research. Cells are not brains, obviously, but new patient-derived stem models have the potential to elucidate cell biological aspects of brain diseases that are not present in worm, fly, or rodent models, the work horses of disease investigations and drug discovery. Neural stem cells are present in the olfactory mucosa, the organ of smell in the nose. Patient-derived olfactory mucosa has demonstrated disease-associated differences in a variety of brain diseases and recently olfactory mucosa stem cells have been generated from patients with schizophrenia, Parkinson's disease, and familial dysautonomia. By comparison with cells from healthy controls, patient-derived olfactory mucosa stem cells show disease-specific alterations in gene expression and cell functions including: a shorter cell cycle and faster proliferation in schizophrenia, oxidative stress in Parkinson's disease, and altered cell migration in familial dysautonomia. Olfactory stem cell cultures thus reveal patient-control differences, even in complex genetic diseases such as schizophrenia and Parkinson's disease, indicating that multiple genes of small effect can converge on shared cell signaling pathways to present as a disease-specific cellular phenotype. Olfactory mucosa stem cells can be maintained in homogeneous cultures that allow robust and repeatable multiwell assays suitable for screening libraries of drug candidate molecules. STEM CELLS2012;30:2361–2365


The pharmaceutical industry is withdrawing from neuroscience research because of the failure of drug candidates to translate from current disease models to success in clinical trials [1]. Modern molecular genetics has been very successful in identifying genes responsible for monogenic diseases and the functions of these genes and their proteins which are then investigated in cellular and animal models. This enterprise is very productive and is the backbone of our deepening understanding of cellular functions. Despite this, molecules successful in correcting the effects of disease-genes in cell, fly, worm, or mouse are not routinely translated easily into clinically useful drugs. The situation is worse for the majority of cases of disease that have multiple unknown genetic and environmental causes [2]. Furthermore, disorders of cognition like schizophrenia become impossible to model exactly in animals because of their unknown origins and complex cognitive and behavioral clinical phenotypes. Even diseases caused by single gene mutations differ in their penetrance and severity in different individuals demonstrating that other genetic, epigenetic, or environmental factors may be operating. One approach is to investigate patient-derived samples because they possess any causative gene mutations as well as the “genetic background” against which the mutated genes act. This approach compares multiple samples derived from patients and healthy controls to identify disease-associated differences in cell biology against a background of individual variability. Human brain tissue is important in this regard but the approach is limited by access to sufficient samples of appropriate quality, both in clinical diagnosis and post-mortem handling [3–6]. Obviously, brain tissue is not useful for dynamic investigations of signaling pathways and drug discovery. Clearly, there is an imperative to develop cellular models that reflect the biological bases of human neurological diseases.commonly, investigators have used accessible non-neural cells such as fibroblasts or transformed lymphocytes (lymphoblastoid cell lines) [7–11]. More recently, induced pluripotent stem (iPS) cells have aroused huge interest and activity in numerous brain diseases because of their ability to be differentiated into specific cells of interest [12]. Our model is based on patient-derived neural stem cells from the olfactory organ, a regenerating part of the nervous system accessible within the nose.


Olfactory function is diminished in many brain diseases, independently of changes with age [13, 14] and post-mortem studies report changes in olfactory mucosa-associated with neurodegenerative diseases, schizophrenia and Rett's syndrome [15–17]. The olfactory mucosa is the sense organ of smell in the nose. The olfactory sensory neurons exist in the olfactory epithelium, the most superficial layer of the mucosa and are replaced through neurogenesis that continues throughout adult life [18–20]. Within the epithelium are basal stem cells that can regenerate this tissue after damage, including the production of new sensory neurons that remake connections to the brain [21–23]. Below the epithelium, within the lamina propria is another stem cell with both mesenchymal and neural characteristics [24–26]. Human olfactory mucosa, obtained by biopsy through the nostril [27], can be grown in tissue and cell culture [20, 27–29]. Using methods derived for mouse neural stem cells, olfactory stem cells can be grown in “neurosphere” cultures. The resulting human olfactory stem cells are multipotent, generating neurons, astrocytes, and oligodendrocytes and propagating through many generations [30, 31]. Neurospheres derived from olfactory mucosa can also generate neural and non-neural cells in vitro, and after transplantation into the chick embryo and into rat models of Parkinson's disease and intervertebral disk injury [30, 32, 33].


“Neuroblasts” capable of continuous culture were developed 20 years ago from olfactory epithelium of human cadavers [28]. Olfactory neuroblasts from patients with Alzheimer's disease had elevated levels of amyloid precursor protein [34], altered concentration of five other proteins [35], and evidence of oxidative damage [36], compared to cells from healthy controls. Neuroblasts derived from biopsies of olfactory epithelium of living patients and controls demonstrated the concordance between fragile X mental retardation one gene and its protein product in neural cells and leukocytes [37]. Olfactory biopsies from patients with schizophrenia were grown in tissue culture and showed more proliferating cells compared to biopsies from healthy controls or patients with bipolar disorder [38, 39]. Biopsies from schizophrenia patients were also less adhesive to cell culture plastic [38]. Collectively, these studies indicate that olfactory cells and tissues can express disease-associated phenotypes but, because the methods used are limiting for extensive and multiple experiments, we developed olfactory stem cell cultures by propagating them initially as neurospheres [30], based on the method for neural stem cell culture from mouse brain biopsies [40]. Primary cultures of cells from olfactory mucosa biopsies are grown in a serum-free medium containing epidermal growth factor and fibroblast growth factor 2 to generate neurospheres that are then dissociated and grown as adherent cultures in standard serum-containing medium [25]. These “olfactory neurosphere-derived” (ONS) cells are a homogeneous population of neural stem cells, but they also express markers of mesenchymal stem cells (CD105 and CD73) as well as other stem and progenitor cell proteins (OCT4, NES, and TUBB3 [25]).

ONS cells are easily produced from olfactory mucosa biopsies; they can self-renew in culture and can differentiate into neurons and glia; and they take approximately 4 weeks to produce working stocks for experimentation. Being able to produce large numbers of homogenous cells, and to bank them, has allowed investigations of cohorts of patient and control cells which embrace individual differences to seek disease-associated variability. We hypothesized that despite differences in genetic (or epigenetic) contributions from different individuals with a brain disease, the causal factors would converge in disease-related signaling pathways that would alter cell functions in a disease-specific manner.

To test this hypothesis, we investigated ONS cell lines from patients and controls in schizophrenia, a highly heritable neuropsychiatric, neurodevelopmental disorder [41], and in Parkinson's disease, a neurodegenerative disease that is heritable only in approximately 5% of familial cases [42]. These are both complex, polygenic diseases of unknown etiology which present significant challenges to cell models because variability between individuals and their cells has the potential to obscure disease-associated differences. We initially investigated 42 ONS cell lines from patients and healthy controls to ask whether they would show disease-associated differences in cell biology and, if so, would these differences be disease-specific. We undertook unbiased gene and protein expression profiling and applied a set of standardized in vitro analyses of cell metabolic functions. These experiments demonstrated that ONS cells from patients have disease-specific alterations in cell biology that define them as different from each other and from healthy controls. In schizophrenia patient-derived cells, 1,700 genes were differentially expressed while in cells from Parkinson's disease patients, 514 genes were differentially expressed compared to cells from healthy controls. The misexpressed genes in schizophrenia were overrepresented in cell signaling pathways associated with neural development, whereas in Parkinson's disease, the misexpressed genes were overrepresented in pathways associated with mitochondrial function and oxidative stress. These pathways are consistent with current hypotheses of the etiologies of these diseases [43, 44], but they were identified without making any a priori assumptions through experimental design. Protein expression analysis identified altered expression of 30 proteins in schizophrenia ONS cells and 11 proteins in Parkinson's disease ONS cells which were consistent with gene expression. Of the six cell function assays, schizophrenia ONS cells were different from control ONS cells in one related to neurodevelopment (“Caspase-3/7 activity”), whereas Parkinson's disease ONS cells differed in two functions related to mitochondrial function and oxidative stress (“MTS (3-(4,5-dimethylthiazol-2-yl)-5-(3-carboxymethoxyphenyl)-2-(4-sulfophenyl)-2H-tetrazolium) metabolism” and “Reduced glutathione content”). These results confirmed our hypothesis that patient-derived ONS cells would differ from control-derived ONS cells in a disease-specific way. Reassuringly, the disease-specific differences were consistent with evidence from a variety of other published studies but, while supporting the literature heuristically, ONS cells revealed specific and previously unknown signaling pathways that provide new leads for investigation.


One of the neurodevelopment-associated pathways that was significantly altered in schizophrenia ONS cells was “G1/S-phase transition” [25]. This is the checkpoint in the cell cycle that regulates whether and when a cell will start DNA synthesis before mitosis and cell division. A finding of a dysregulated cell cycle is of interest because we had previously noted that in olfactory mucosa biopsy cultures, there were more cells in mitosis in schizophrenia tissue than in tissue from healthy controls [38, 39]. We therefore investigated cell proliferation and cell cycle dynamics in ONS cells [45]. Schizophrenia patient-derived ONS cells proliferated more rapidly than control ONS cells and this was associated with: (a) a larger fraction of the cells entering the cell cycle after holding them in G0/G1 phase; (b) a 2-hour shorter cell cycle period in the dividing cells; and (c) higher levels of the protein cyclin D1, the regulator of G1/S-phase transition, as well as (d) raised levels of cyclin E and cyclin A2, whose levels are triggered in a cascade as the cell moves from G1 to S-phase. These observations indicate that, in schizophrenia, there may be alterations in the signaling pathways regulating the cell cycle. We do not know what is driving the higher cyclin D1 levels in patient cells, but this is the final common node that controls the rate of the cell cycle upon which multiple upstream signaling events converge. These provide new targets for investigations.


One of the most affected signaling pathways in the ONS cells from patients with Parkinson's disease was “NRF2 signaling,” a pathway that acts on oxidative stress [25]. Experiments in the fruit fly suggested that NRF2 may be a useful therapeutic target in Parkinson's disease [46]. We investigated the involvement of NRF2 signaling functions in Parkinson's disease patient-derived ONS cells [47]. NRF2, a transcription factor activated by oxidative stress, leads to transcription of several genes that ameliorate the effects of oxidative stress. Patient cells and control cells had similar levels of NRF2 but patient cells had lower levels of its downstream targets. Hence, NRF2 activation was reduced in patient cells. Accordingly, we activated NRF2 with L-sulforophane in these cells and were able to increase to control levels, their reduced glutathione and MTS metabolism. Conversely, in control cells, we decreased assays to patient cell levels by blocking NRF2 translation with siRNA. These observations demonstrated that patient cells are under oxidative stress under normal culture conditions, partially because their NRF2 signaling pathway is impaired. It is likely that upstream regulators of NRF2 signaling are responsible because although L-sulforophane treatment restored cellular functions in patient cells, gene expression remained significantly different from control cells, indicating the NRF2 signaling is probably not the primary regulator of homeostasis in these cells in vitro.


Homeostatic regulation maintains all functions of cells within narrow bands to keep cells viable according to external environmental and internal states. Homeostasis will obviously be a complex interaction between multiple processes (transcription, RNA processing, translation, post-translational modification, ubiquitinylation, etc.) and the operation of the resultant signaling pathways. Gene expression profiling of cells in vitro provides insight into one facet of this regulatory state of homeostasis. Expressed transcripts will include regulators of homeostasis as well as regulated genes, such that expression profiles between individuals will reflect individual differences as well as disease-associated differences. The absolute level of gene expression in an individual cell may reflect the homeostatic drive affecting that gene within the regulated landscape. Conversely, different cells or individuals may differ in the absolute level of gene expression, reflecting individual differences in homeostatic drive. This variance in gene expression may affect the outcomes of cell functions in response to their environment, such as the final outcome of cell differentiation and maturation suggested by Waddington's canalized genetic and epigenetic landscape [48]. We hypothesized that there may be disease-associated differences in gene expression variance associated with the gene expression profiles of the 42 ONS cell lines from patients with schizophrenia or Parkinson's disease, and healthy controls [25]. We investigated the variance of gene expression in cell signaling pathways that were most active in the ONS cells and found that (a) in schizophrenia patient ONS cells, gene expression was less variable and (b) in Parkinson's disease patient ONS cells, gene expression was more variable than in control ONS cells [49]. The variability in gene expression was concentrated in cell surface receptors and signaling molecules. These observations suggest a novel concept in disease-associated differences in cell biology, namely that variability of gene expression (not just absolute levels) may contribute to individual differences in the highly regulated networks of homeostatic regulation.


Olfactory stem cells exist in primary cultures of the olfactory mucosa [24] from which ONS cells are derived [25]. In patient-derived primary cultures of olfactory cells, there were significant differences from control cells in familial dysautonomia, a recessive monogenic disease caused by mutations in the IKBKAP gene [50]. Patient cells expressed mis-spliced transcripts and lower levels of the normal transcript of this gene and its protein. Gene expression profiling identified that patient cells had altered expression of genes associated with the cell signaling pathways of cell migration and cytoskeleton reorganization. Furthermore, patient cell migration was impaired compared to control cells [50]. It was additionally demonstrated that the cytokinin, kinetin, was able to correct IKBKAP mRNA splicing to increase the expression of IKBKAP protein [51].


ONS cells are not the only promising patient-derived stem cell models of brain diseases: iPS cells [52] are also well-advanced for this purpose [12]. iPS cells have been generated from patients with familial forms of amyotrophic lateral sclerosis [53], muscular dystrophy, Huntington's disease, Gaucher disease, Down's syndrome [54], spinal muscular atrophy [55], dysautonomia [56], Parkinson's disease [54, 57, 58], and schizophrenia [59, 60]. Most of these patients have monogenic diseases; monogenic diseases may be more tractable for cell models because functional effects in patient-derived cells can be assumed to be downstream of the causative mutation. At a practical level, monogenic diseases may also be preferable for iPS cell analyses because investigations can target downstream effects of the causative mutation in small numbers of patients and controls. By contrast, for genetically complex diseases, it will be necessary to compare cells from large numbers of patients. For example, iPS cell lines were generated from five cases of sporadic Parkinson's disease and differentiated into neurons but no differences were observed between patient and control cells [58]. Conversely, in schizophrenia, which is more heritable, a study of iPS cells from three related cases identified several disease-associated alterations in neuronal differentiation including reduced glutamate receptor expression and reduced interconnectivity between neurons in vitro [59].

A property of iPS cells is the relative ease with which they can be differentiated into neurons, but their wide application as models of brain diseases faces significant technical challenges [61], not the least being the cost and time required to establish and maintain each cell line, which may inhibit the ability to compare iPS cells from multiple patients. For patient-control comparisons, it will be important to distinguish disease-related variability from technical variability arising from differences between individuals, clones, and the reprogramming process. As the reprogramming process becomes better understood, and cell reprogramming technology improves, some of these issues may be resolved and the time and cost may be reduced. For example, as it is now possible to generate neurons directly from skin fibroblasts [62], the cost of producing patient-derived neurons may decrease because fibroblast cultures are much cheaper to generate, expand, and maintain than iPS cells.

Modeling complex brain diseases with patient-derived cell models faces the fundamental challenge of whether cell populations in vitro can recapitulate disease processes in the brain that may take years to develop or may involve multiple cell types [12, 61]. “Disease modeling” using patient-derived cells is aimed at gaining insight into the cellular and molecular dysfunctions that underlie brain pathology, rather than aiming to simulate functions of the diseased brain. Unraveling disease processes will require careful selection of control groups, to reduce the effects of potential confounds such as medication exposure, age, and duration of illness. Additionally, for each disease, the requirements of the patient-derived cell “model” vary and researchers may choose to focus on different aspects. For example, in schizophrenia, a neurodevelopmental disease, ONS cells were used to investigate neurogenesis and neural development [25, 45], whereas iPS cells were used to investigate aspects of neuronal connectivity [59]. In idiopathic Parkinson's disease, a neurodegenerative disease, patient-derived ONS cells demonstrated dysfunctional regulation of oxidative stress [25, 47], consistent with dopaminergic neurons derived from iPS cells from patient with a familial form [57] but not patients with idiopathic forms of the disease [58]. In familial dysautonomia, there is significant convergence in disease-associated cell pathology in patient-derived olfactory “ectomesenchymal cells” and in “neural crest cells” differentiated from patient-derived iPS cells: mis-splicing of the IKBKAP gene, reduced levels of IKBKAP protein, defects in cell migration, and correction of the mis-splicing by kinetin treatment [50, 56].

ONS cells make very practical models for brain diseases because they are not genetically modified and they can be generated quickly in large numbers to produce homogeneous populations for functional assays. Predictable, repeatable cell populations are a requirement for high throughput screening for drug discovery. ONS cells fulfil this criterion and we have shown that patient-control differences in response to drug-like molecules can be measured in multiwell plate assays [47], another criterion for practical screening of large libraries of candidate molecules. We envisage ONS cells will play a role in discovering small molecules that alter disease-associated cell functions with application as probes of cell biology and as future drug candidates. Many lead compounds discovered in cell-free assays and in immortalized cell lines subsequently fail toxicity tests. ONS cell screening would simultaneously deselect molecules toxic to human cells. It is expected that the range of applicable cell functions will increase as methods are developed for ONS cells to differentiate them into desired neuronal subtypes.


This work was supported by a grant from the Australian Government Department of Health and Aging to establish the National Centre for Adult Stem Cell Research and by project grants from the National Health and Medical Research Council of Australia.